In the 0.6 series, the aptitude resolver is a lot more conservative
than it used to be, due to repeated complaints by users that it was
removing too many packages in the first solution that was presented.
It looks like maybe I overcompensated, though (this is not the first
report along these lines).

  The main difficulty in cases like this is getting the algorithm to
properly balance removals with upgrades.  However, it looks like the
old scoring system works just fine here.  This command line:

aptitude -o 'aptitude::problemresolver::remove-tier=10000' -s dist-upgrade

  tells aptitude that it shouldn't force removals to appear after
keeps.  The resulting solution looks mostly reasonable:

      Remove the following packages:                       
1)      debian-keyring                                     
2)      flightgear                                         
3)      java-gcj-compat                                    
4)      java-gcj-compat-headless                           
5)      libghc6-parsec-dev                                 
6)      libghc6-parsec-prof                                
7)      supertuxkart                                       
8)      texlive-base-bin                                   

      Keep the following packages at their current version:
9)      libplib1 [Not Installed]                           
10)     simgear1.9.1 [Not Installed]                       

      Leave the following dependencies unresolved:         
11)     devscripts recommends debian-keyring           

  It still isn't perfect, but that seems mostly due to problems with the
flightgear package and its dependencies.


  I have some ideas on how to use the tier mechanism in a more flexible
way to balance keeps with removals, but I don't know if I'll get to
implement them before Squeeze (after all, the freeze is next month).  It
might be simpler to just change the defaults to the settings I gave
above, so that upgrading to Squeeze works as expected.

  Daniel



-- 
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to